Performing Measurements

Avoid DSP Configuration Errors That Can Invalidate Your Work

Before performing any measurements, it's essential to verify that your DSP device is in a clean reset state. Otherwise, the final results after performing the work of measurements and optimizations may contain errors that render all that work invalid. Three possible concerns are described below.

Clear Out DSP Filters, Delays, Gains and Polarity Inversions Before Measuring

Before measuring, all system EQ should be disabled, both in your AVR and any external DSP hardware such as miniDSP devices. This includes AVR Dynamic EQ.

It's important to realize that the filter parameters, polarity inversions and delay and gain values used by MSO represent changes relative to the conditions at which the measurements were performed. If MSO recommends a delay of e.g. 7.4 msec for a particular sub, and the measurements were performed with a non-zero delay set for that sub in the DSP device (let's say 2 msec), the correct final delay for that sub would not be 7.4 msec, but 9.4 msec in this specific case. It should be clear that performing the measurements with any delay set to a non-zero value has the potential for confusion and error. To avoid such problems, it's much easier to simply zero out all DSP delay and gain/attenuation parameters prior to performing the measurements.

A similar, but even worse situation can occur if the measurements were performed with DSP filters set to a non-flat condition. In the case of miniDSP hardware, exporting a biquad text file from MSO and importing it into a DSP channel that already had some filters enabled during the measurements will usually cause those filters to be overwritten, not appended to. This guarantees an incorrect result, not just confusion and inconvenience as with the above delay example.

Important Exceptions to the "Completely Clear Out the DSP" Rule

In certain special situations, you may need to make an exception to the above rule of measuring with the DSP completely cleared out. To find out when this might be necessary, see the Making Exceptions to the Rule of Measuring With the DSP Completely Cleared Out subtopic in the Tech Topics section.

Verify Signal Routing For miniDSP Devices

Most miniDSP devices used with MSO have a so-called input routing matrix controlling how the inputs are connected to various outputs. Below is a summary of how the routing should be configured for the popular miniDSP 2x4 HD device. Other devices should be similar.

MSO does not support using multiple subwoofer outputs from your AVR, so your DSP device must be configured to use only one input.

Usage of Third-Party Programs For miniDSP Configuration May Require a Reset

If you've been using any third-party software for configuring a miniDSP device, you may need to reset the device before doing any data entry into the miniDSP software. This is to ensure that all the device settings are consistent with their display in the software. Such third-party software, when used before running the factory miniDSP software, can cause the factory miniDSP software to display or set an incorrect state. See The miniDSP "Write-Only" Problem in the Tech Topics section for detailed information about using the reset to keep the miniDSP software in sync with the actual device state.

Avoid Measurement Errors That Can Invalidate Your Work

As mentioned in the previous page's warning, neglecting certain measurement requirements can lead to invalid results at the end of the process. This section aims to provide details of the pitfalls that can arise and how to avoid them.

Imported Measurements Must Use a Timing Reference

MSO requires that the measurements you import be time-synchronized with one another. This means that the relative delays (and therefore phase relationships) between all subs and loudspeakers measured at a given listening position will be preserved in the measurements. This feature is not automatic in measurement software. It requires special attention in order to enable it. Ensuring this phase-preserving property involves the use of a timing reference.

Using an Acoustic Timing Reference With a USB Microphone

When using a microphone with USB output, the acoustic timing reference of Room EQ Wizard must be used. When using an acoustic timing reference, the same speaker must be used as a timing reference for all measurements. This speaker must be one with a high-frequency driver that can properly reproduce a sweep whose lowest frequency is 5 kHz. A subwoofer cannot be used for this purpose.

REW Acoustic Timing Reference Considerations

For instructions for how to do REW measurements with an acoustic timing reference, see the REW documentation on the topic.

Use the REW Adjust Clock With Acoustic Ref Option

Using the REW Adjust Clock With Acoustic Ref option in its Analysis Preferences dialog is strongly recommended. This option may improve the correlation between MSO's predictions and the final results you measure with all subs energized at once. It can also pick up cases for which the sample rate chosen in Windows for either the microphone (recording device) or the AVR (playback device) is incorrect. Some guidelines for these settings follow.

Before measuring in REW when using Windows, always double-check the Windows sample rates discussed above, as Windows has a tendency to reset them as you disconnect and reconnect devices.

If these guidelines aren't followed, Windows may resample the audio, potentially causing phase inaccuracies that affect the correlation of MSO's predictions and the final results you measure with all subs energized at once.

The REW Adjust Clock With Acoustic Ref option in its Analysis Preferences dialog is only designed to compensate for minute hardware clock differences when the clocks of the recording and playback devices are nominally equal. It's not designed to fix sample rates that are incorrectly chosen in software.

Verifying Your Timing Reference

To make sure your timing reference is working properly before performing a full set of measurements, you should run the two-sub timing reference test described on the previous page.

Using a Timing Reference With an XLR Microphone (Uncommon)

To obtain time-synchronized measurements with REW and an XLR microphone such as the Parts Express EMM-6, you must use a separate sound device - usually an external USB sound device - with a microphone input. Measurements may be performed using either a loopback (hardware) timing reference if your hardware supports it, or an acoustic timing reference. Loopback timing references are tricky when there is a device with a digital input, such as an AVR, in the measurement signal path. Ask in the forums if you need help with this.

XLR-to-USB Adapters

When using an XLR microphone in conjunction with a so-called "XLR to USB adapter", this "adapter" is really an external USB sound device. Therefore such a combination should be considered as the equivalent of a USB microphone, and can only be used in conjunction with an acoustic timing reference to achieve time-synchronized measurements.

The Measured Sub Delays in REW Are Not Meaningful to MSO

This note in the REW documentation about its measured sub delays is worth repeating.

"For speakers the delay estimate is based on the location of the peak of the impulse response. Subwoofers have a broad peak and a delayed response due to their limited bandwidth so the delay is instead measured relative to the start of the impulse response. The start of the impulse response cannot be located as precisely as the peak, however, so delay values are less accurate for subwoofer measurements."

MSO uses only frequency-domain data you export from REW and ignores the delay values REW computes from the time-domain data. This means you can ignore REW's calculated system delay values, including subwoofer delays.

Do Not Shift Any Impulse Responses in REW

MSO requires that the actual relative delays between all loudspeakers measured at a given listening position, including mains and subs, be preserved in the measurements. Even if a timing reference is used in the REW measurements, certain manipulations you can do in REW will shift the impulse responses and break these relationships. These include the following:

If you perform either of these operations before exporting the data, you must redo all the measurements. If you wish to do such experiments in REW, it's best to make a copy of your .mdat file and do the operations on the copy.

Smoothing Should be Disabled When Exporting REW Measurements as Text

It is essential that the measurements exported from REW and imported to MSO be unsmoothed. Otherwise, errors will be introduced between MSO's predictions and the final measured responses. Removal of smoothing is most easily accomplished when exporting the data from REW. In REW, choose File, Export, Export all measurements as text. This will cause the dialog below to be shown.

REW's Measurement Export Dialog
REW Measurement Export Dialog

Make sure Use custom smoothing is selected as above, and its combo box set to No smoothing.

Measurement Precautions to Optimize Your Final Result

The methods described below will help you to avoid unnecessary degradations of final performance after optimization that can happen when best practices for measurement aren't followed. Following these guidelines will help you achieve the best possible result after optimization.

Sub-Only Optimization: Prevent High-Frequency Roll-Off From Undermining Your Results

When optimizing a sub-only configuration, you should avoid conditions under which high-frequency roll-off occurs in the measurements. The reasons why this is a problem will be shown in detail in the sections following this one. But first, we'll look at some methods to avoid it.

The Causes of High-Frequency Roll-Off In Sub Measurements

It's useful to separate the causes of high-frequency roll-off in sub measurements into different categories: that which is caused by the AVR and that which is caused by the sub's amplifier or the sub itself.

Roll-Off Caused by the AVR

Traditionally, there have been two conventional ways to energize your subs when performing measurements to be used with a sub-only configuration.

Dealing With AVRs Whose "LPF for LFE" Cannot Be Adjusted

For some AVRs and pre-pros, the "LPF for LFE" is not adjustable, but fixed at a common value, often 120 Hz. For such hardware, the LFE channel (HDMI channel 4) must not be used to energize the subs. Instead, energize the subs through the bass management, or by using the technique described below in A Trick for Eliminating All Low-Pass Filters from the AVR's Signal Path.

A Trick for Eliminating All Low-Pass Filters from the AVR's Signal Path

If you want to eliminate all low-pass filters from the AVR's signal path when measuring the subs, there is a trick that can be used. This trick has become more popular over the years, and is recommended for measurements used with MSO sub-only configurations.

Roll-Off Caused by the Sub Amplifier

High-frequency roll-off of sub responses due to unwanted low-pass filters in the signal path can occur in active subs and even in systems using passive subs when pro audio amplifiers are used.

Roll-Off in Active Subs

Be careful of the amplifier settings of active subs. Not only can sub amps cause undesired high-frequency roll-off of the measurements, but the techniques for eliminating it can be confusing or, unfortunately in some cases, non-existent.

If the subs are of the active type, they may have their own built-in crossover. This feature is intended for two-channel users whose electronics rarely include such a crossover. This crossover should be bypassed completely if at all possible. If it's not possible to disable it, the crossover frequency must be set to the highest value possible. Sometimes the ability to disable the crossover is provided using the knob that sets its frequency. There might be a detent on this knob, labeled "Out" or similar. If such a feature is available, use it.

Active subs sometimes have an input without a crossover, often somewhat confusingly called "LFE". This input is intended for use with AV equipment already having a crossover (all AVRs and pre-pros). This input should be used if available. Certain sub models, such as some from SVS, use their LFE input for multiple purposes. For such subs, forcing the highest sub bandwidth possible might require an additional step of setting the sub to "LFE mode".

See your subwoofer's documentation for details of these settings.

Roll-Off in Passive Sub Systems Using Pro Audio Amplifiers

Pro audio amplifiers with built-in DSP such as the Behringer NX6000D have become popular with DIY subwoofer builders. If the DSP of such an amp is being used, make sure it's configured to not use any low-pass filters.

Even some non-DSP pro audio amplifiers can put unwanted low-pass filters into the signal path if not configured correctly. The popular Behringer NX6000 (non-"D" version without DSP) has a switch on the back labeled CROSSOVER that can cause this problem. Here's an excerpt from its user manual describing this switch.

"[The] CROSSOVER switch chooses between three modes: FULLRANGE, LF (low frequency crossover) and HF (high frequency crossover). In LF mode, the unit amplifies only the low frequencies of the signal. In HF mode, the unit only amplifies the high frequencies. LF and HF modes are typically used in bi-amping applications."

When using this amplifier to drive a subwoofer, the CROSSOVER switch should be set to the FULLRANGE mode, not the LF mode. The LF mode will introduce an undesired low-pass filter into the signal path.

Roll-Off Caused by the Sub Itself

An underappreciated cause of high-frequency roll-off of subs is the voice coil inductance. Generally, this effect is most pronounced in some of the large, high-output, high-excursion subs favored by DIY subwoofer enthusiasts. One example that stands out is the Dayton Audio HTS545HE-4 21" driver. Unfortunately, the effect of voice coil inductance is difficult to predict in simulation software using the manufacturer's data for the following reasons:

What's needed is to have a simulation program for subs that's easy to use like WinISD, but is able to use the advanced inductance parameters described by Thorborg. Also, Dayton DATS, which is easy to use because of its pre-built fixtures and software, should be updated to be able to extract these parameters.

The best way to figure out the impact of driver voice coil inductance on high-frequency roll-off as of this writing is to do a near-field acoustic measurement with REW, using flat-gain electronics to drive the subs.

Why High-Frequency Roll-Off Can Undermine Sub-Only Optimization

So far, we've talked about high-frequency roll-off and how to avoid it, but the rationale for the claim that it can hurt our final performance hasn't yet been discussed.

To see why it can be a problem, let's go back to an earlier example in which we optimized SPL without trying to shape the response. After that step, we had to set the reference level for an anticipated later optimization to flatten the MLP response.

The graph from that example is replicated below.

The MLP After Optimizing SPL
The MLP After Optimizing SPL

There's some roll-off above 100 Hz that's disappointing. We'll assume for now that we can't get rid of it.

Normal Response Level

We have a large peak of 25-30 dB at around 25 Hz. That's not a problem, because the room mode that causes it is giving us a response boost above what might be called the "normal level". This "normal level" runs from about 50 Hz to about 120 Hz in the figure above. The boost above the normal level results in an SPL that's much higher than what we'd get in an open-air environment without room modes. This response boost allows us to reach our SPL target with less amplifier power and subwoofer displacement than if it were not present.

In this example, as we did earlier, we can picture an imaginary horizontal line at a level of 80 dB. Using only cuts and no boost, we can scale the response down to this 80 dB level in the figure above so it's flat from 10 Hz to almost 200 Hz. Doing the equalization in this way means that, by using only PEQ cuts, we're knocking the response down to a level that's at the bottom of the troughs of the response in the frequency range where the response is in the region of what we called the "normal level" above. If we were to pick a level lower than 80 dB, we'd need correspondingly more cut in the equalization to reach our target.

This type of equalization is a frequency-dependent gain cut. To compensate for it, an increase in sub gain is usually required. This is fine in moderation, but if excessive gain cuts are implemented in equalization, there may not be enough sub gain available to make up for it. Unwanted LPFs in the subwoofer amplifier can cause exactly this problem. We'll see in the next section how that can happen.

The Effect of an Unwanted Sub LPF

Suppose that our sub amp has an LR4 crossover at 120 Hz that we can't get rid of. We can simulate the presence of this filter in MSO just to illustrate. We'll skip the details of how this is done for now. After adding this undesired LPF, we get the following result.

The MLP With an Unwanted 120 Hz Sub Crossover
The MLP With an Unwanted 120 Hz Sub Crossover

We're getting in trouble now. The prior 80 dB reference level intersects the new curve at about 117 Hz, rather than just below 200 Hz as it did before. If we want our equalized response to be flat up to one octave above the crossover, this problem would limit our maximum crossover frequency to 60 Hz at best. Not only that, but the 120 Hz LPF introduces a frequency-dependent phase lag that will cause difficulties when integrating main speakers and subs later. That's why it's essential to bypass any LPF that might be present in the amplifier of an active sub.

If we wanted to extend our response out almost to 200 Hz again, we'd need to pick a reference level of 65 dB or less in the figure above rather than the original 80 dB. That means the response must be knocked down by an additional 15 dB so that it reaches a level of 65 dB at all frequencies, just to match the level set by some high-frequency roll-off that shouldn't be there in the first place. That's not practical or reasonable, as we'd lose another 15 dB of gain in a system that already had a substantial gain cut. That gain cut is partly traceable to the roll-off above 100 Hz in the original data. In an ideal situation, we'd like to get rid of that too. If we've exhausted all possibilities of fixing the source of the roll-off, we can use an HF shelving filter to counteract it. That technique was demonstrated earlier in the example for flattening the MLP response.

General Measurement Considerations

The following guidelines can help with planning before taking your actual measurements.

Measure the MLP Last to Help With Verification

You can help make the later task of verifying MSO's performance predictions against the final measured REW data easier by following the simple step below.

If your domestic situation allows it, measure the MLP last and keep the microphone in place for when you do the final verification. Using this technique will all but eliminate the measurement variability at the MLP caused by differences in microphone location between the original measurements and the final verification. This is described in more detail in the FAQ.

What Measurements Do I Need to Take?

For each listening position you wish to measure, you need to measure the frequency response of each sub individually at that position. If you are using MSO to integrate the mains and subs (uncommon nowadays), you must also measure the main speaker(s) at each listening position. Suppose you are measuring Ns subs and Nm main speakers at Np listening positions. This means the number of measurements you must perform is (Ns + Nm) * Np.

Example: If you're optimizing a sub-only configuration, then Nm is zero. If you are optimizing, say, 4 subs at 5 listening positions, Ns = 4 and Np = 5. The number of measurements is then (Ns + Nm) * Np = (4 + 0) * 5 = 20.

How Many Positions Should Be Measured?

This depends on your setup. In general, the number of positions measured should be the number of seats you're concerned about. If you are more concerned about some seats than others, you can specify error weighting on the Group Weights Property Page. Starting with MSO v2, application of this error weighting has been expanded to cover all optimization types listed on the Optimization Type property page of the Optimization Options property sheet except Flatten MLP response using only shared (input) filters.

Should I Measure at Different Heights?

If you have subs at different heights, and these subs are on separate DSP channels, it's possible to reduce the effect of height modes by incorporating measurements taken at different heights into the measurements you use for MSO. Stacked subs with the same signal applied to them don't count in this regard.

If you measure at different heights, but don't have independently-controlled subs at different heights, this won't harm anything, but it won't allow for reducing the effect of height modes either.

Subwoofer Measurement Considerations

When performing subwoofer measurements for use with MSO, it's essential that only one subwoofer and no other loudspeaker be energized at a time. Assuming you have a DSP device for performing individual EQ of each sub, this means muting all outputs of the DSP device except the one associated with the sub being measured. If you have more subs than you have DSP output channels, then more than one sub will need to be connected to a given DSP output channel via a Y cable or similar. Such "doubled up" subs count as a single "logical sub" for the purposes of MSO, and should be measured together rather than individually.

Measuring With Room EQ Wizard (REW)

REW is a popular choice for measurement software, and is recommended as the first choice on all platforms for performing measurements for use with MSO.

There have been many recent changes to REW. Version 5.30 in particular saw major updates. See What's new in REW V5.30 to find out what changed. To prevent confusion in forum discussions, it's a good idea to upgrade REW to the latest non-beta version for doing your measurements.

Some Important REW Documentation

In the Getting Set Up for Measuring section of the REW documentation, the following topics are especially important.

Also, be sure to review the documentation for the Soundcard Preferences dialog.

For Windows REW Users

Multi-Channel Measurement Considerations

If you're using a Windows machine for your measurements and applying a stimulus to any channel other than Left or Right, you'll need to know about the multi-channel measurement methods available to you.

In earlier versions of REW running Windows, the only way to perform measurements using device channels other than Left or Right was to use the ASIO interface. ASIO is an application programming interface (API) originally defined in a specification by the sound device manufacturer Steinberg. Any installed software that implements the ASIO API correctly will be available as a choice in the REW Soundcard Preferences dialog. There are two categories of ASIO implementation: ASIO wrappers and native ASIO drivers.

ASIO Wrappers

An ASIO wrapper is software that implements the ASIO API using already-existing Windows APIs for controlling sound devices. There are two common ones.

ASIO wrappers allow you to choose different input and output devices in REW. This is essential when measuring using an AVR, which requires that the REW output be set to HDMI and its input typically to a USB microphone.

A good tutorial for Windows users making use of an ASIO wrapper is AustinJerry's guide (in PDF form), called "Getting Started With REW: A Step-by-Step Guide". It's a good resource for those doing measurements with HDMI and a USB microphone (which requires an acoustic timing reference). This document is very detailed, so if you're having trouble, you can go through it step by step to get up and running. This guide predates the availability of the WASAPI Exclusive devices introduced in REW V5.30.

Native ASIO Drivers

A native ASIO driver is software that implements the ASIO API by communicating directly with the sound device hardware, bypassing the Windows APIs normally used for controlling sound devices. Such drivers only work for the specific sound device they're written for. These are common in pro-style sound devices made for music recording due to their low latency, which is critical for such applications.

Native ASIO drivers do not allow you to choose different input and output devices in REW. They are only useful in situations where you're applying an analog signal to the device under test via the sound device, and measuring acoustically using an analog microphone (usually with an XLR connection) connected to a microphone preamp on the same sound device.

New WASAPI Exclusive Devices in REW

Prior to REW version 5.30, the only way to do multi-channel measurements was to use ASIO. The most recent REW versions now make use of the csjsound-provider software by Pavel Hofman. This software allows using sound devices in WASAPI Exclusive mode, which enables multi-channel measurements on devices that support it. According to the REW documentation:

"Devices using WASAPI exclusive are prefixed by EXCL: in REW's device selectors. Note that WASAPI Exclusive is only available for devices that have the "Allow applications to take exclusive control of this device" option selected in the Windows audio properties for the device."

The figure below illustrates the selection of this option in the sound device properties.

Configuring the Sound Device for Exclusive Mode
Configuring the Sound Device for Exclusive Mode

The WASAPI Exclusive device type is a promising new development in REW.

Channel Numbers vs. Physical Channels in Windows

The table below shows the Windows channel numbers for use with REW in a 7.1 setup. For multi-channel measurements, use one of the device types described above that support multi-channel usage.

Channel Number Channel Name
1 Left
2 Right
3 Center
4 LFE
5 Rear Left
6 Rear Right
7 Side Left
8 Side Right

Channel Number vs. Physical Channel for Windows 7.1 Audio

Windows Sound Device Settings

Sometimes, undesired sample rate conversion can occur in the OS without you knowing it. This situation could adversely affect the timing reference accuracy and must be avoided. You can avoid it by appropriately configuring your sound device settings in Windows. Three conditions must be satisfied.

As a specific example, the popular miniDSP UMIK-1 measurement microphone always runs at a sample rate of 48 kHz. This means you must also set the device properties for the microphone and the output device in the Windows sound device Control Panel as follows.

In addition to all that, make sure to set the device sample rates in REW to match. From the REW docs (emphasis in original):

"If WASAPI exclusive is not being used make sure the input and output devices are configured in the OS to operate at the rate selected in REW, otherwise the OS will resample between the selected rate and the rate at which the input or output device is actually running."

This can of course be achieved by setting the sample rate first in Windows, then ensuring the sample rate chosen in REW matches what you set in Windows.

Avoiding the Need for Multi-Channel Measurements

If you're measuring for a sub-only configuration and you use the technique described above in A Trick for Eliminating All Low-Pass Filters from the AVR's Signal Path, you'll only need to use the Left and Right channels in REW: one channel for the timing reference and the other for the subs. This would allow you to use the legacy JavaSound drivers with REW, which could come in handy in case of trouble with the other driver types.

For Mac REW Users

Mac users should check out the Mac REW tutorial thread on AVS forum by Enrico Castagnetti from Rythmik Audio.

Measurement Examples for Subs+Mains Configurations

In some discussion above, the potential pitfalls of performing sub measurements using the AVR's bass management and the acoustic timing reference were described. These potential problems were identified in the context of gathering measurements for sub-only configurations. Using that technique was discouraged because of its added complexity. Unfortunately, performing sub measurements using the AVR's bass management is required for subs+mains configurations, as both the subs and main speakers need to be measured with the crossover enabled, and its frequency set to the anticipated final value.

Because of the potential for confusion, two examples of sub and mains measurements using the AVR's bass management and an acoustic timing reference are provided below.

The example below demonstrates the use of the acoustic timing reference when measuring a system having a Left channel, Right channel, and four subs. It's essential that you use the same channel as the acoustic timing reference for all measurements. The example below uses the right channel for this purpose. Notice that for the subwoofer measurements, the REW output is set to the left channel, but the left speaker is muted or disconnected, so that no sound comes from it. In this case, the left channel energizes the subs through the AVR's bass management. These measurements need to be repeated at each listening position.

Left Channel, Right Channel and Subs

In the optimization calculations for the arrangement above, MSO will detect that there are two main speaker measurements at that listening position, and will boost all subwoofer data by 6 dB before summation with the mains to take into account the fact that both main speakers are summed together, but the subs were only energized by a single main channel (not both main channels simultaneously).

The following example is for a center channel and four subs, using the right channel as the acoustic timing reference. These measurements need to be repeated at each listening position.

Center Channel and Subs